Deep Learning-Based Blind Estimation of Symbol Rate

Blind symbol rate estimation is crucial for adaptive communication systems and cognitive radios. This paper presents a novel approach to tackle blind symbol rate estimation from received complex baseband signals. The method entails the extraction of both the in-phase (I) and quadrature (Q) components from the signal and leveraging a residual convolutional neural network (ResNet) to learn latent features from raw IQ components for symbol rate prediction. Simulations are conducted to assess the effectiveness of the proposed method. The results demonstrate that our IQ-based approach outperforms an existing deep learning-based symbol rate estimation method across diverse conditions, including both additive white Gaussian noise (AWGN) and Rayleigh channels.